To enforce social distancing protocol in public spaces and workplaces, this project intends to build a Social Distance Analyzer AI Tool that can monitor if individuals are keeping a safe distance from each other by analysing real-time video feeds from the camera. It has the ability to identify where each individual is in real-time and produce a bounding box that turns red if the distance between two persons is too close. It can also determine whether a person is wearing a mask or not by utilising a pre-trained CNN-based model. This may be utilised by concerned authorities to analyse people's movements and inform them if the situation worsens.
★ Person detection using YoloV3 and getting bounding boxes in real-time
★ Face detection i.e Extracting Region of Interest (face) in each detected person’s bounding box
★ Face Mask Classifier (MobileNetV2 Architecture) to identify whether person is wearing mask or not
★ Social distancing measurement i.e computing the pairwise distances between all detected people and based on these distances, check if any two people are less than N pixels apart.
★ Display output status in the frame